Abstract
Climate change-driven increases in drought risk pose a critical threat to global carbon and water cycles. However, ecosystem-scale responses remain poorly quantified, particularly for severe, multiyear drought events. We addressed this gap by examining ecosystem-scale carbon and water flux sensitivity to the extreme 2018-19 drought in Australia using data from 14 eddy covariance flux sites. The ecosystems span grasslands and semi-arid woodlands to tropical and temperate forests. The driest sites (classed as "grass" and "very dry") experienced drastic productivity impacts, with a 65% decrease in Gross Primary Productivity (GPP) over 2 years relative to the pre-drought average. However, fluxes in "dry," "seasonally wet" and "wet" ecosystems showed remarkable resistance, with no overall change in GPP. All sites recovered rapidly; carbon fluxes in the first post-drought year matched (and generally exceeded) those of a climatically similar pre-drought year. Drought responses were strongly mediated by ecosystem-specific strategies. The driest ecosystems showed direct coupling of productivity to water availability, while intermediate ecosystems (dry and seasonally wet) leveraged stored soil water to maintain evapotranspiration and productivity under drought. At these sites, water was conserved over wet periods (evapotranspiration < demand, despite sufficient rainfall) and consumed over dry periods (evapotranspiration > rainfall). This mechanism mitigating periodic water stress under high rainfall variability likely contributed to the notable drought resistance of the dry and seasonally wet sites. The monthly water deficit index (MWDI) emerged as a robust predictor of productivity across space, highlighting that short-term water availability deficits strongly influence overall ecosystem composition. Analysis of drought response mechanisms suggested rapid leaf loss under water stress, particularly at the driest sites. Our findings underscore the importance of accounting for sub-surface water storage and diverse drought response strategies in vegetation models. We provide critical benchmarks for improving parameterization of plant-water relations, aiding efforts to inform climate-robust management strategies.